recurrent-interface-network-pytorch

recurrent-interface-network-pytorch copied to clipboard

ValueError: num_samples should be a positive integer value, but got num_samples=0

1

When running the Trainer example script from readme, the program fails to run with the following error: `Traceback (most recent call last): File "RIN.py", line 20, in trainer = Trainer(...

ad48hp

Hi Phil, I wanted to know about the specific compute that you used or some approximate estimate of compute required to get a minimal architecture runnning. Thanks

kushalj001

mixed precision training

13

Hey, I've worked on reimplementing RIN based on the authors repo and this repo but I cannot manage to make it work with mixed precision and I see you do...

nicolas-dufour

Inconsistent interpretation of model output between self-conditioning step and prediction step

1

I think there is a bug in the interface self-conditioning in rin_pytorch.py. The model output is interpreted differently during the self-conditioning stage compared to the prediction stage. Currently we have...

jsternabsci

Hi, Just another suggestion. Wouldn't the more accurate name for the classes Attention and LinearAttention in rin_pytorch.py be CrossAttention and SelfAttention?

katop1234

Hi, I was looking to implement the RIN architecture using your code. I noticed you applied positional embedding in every block, as opposed to just once in the beginning. What...

katop1234

follow up paper

7

recently ran into a researcher who told me there is a follow up paper to this work does anyone know of it?

lucidrains